2026人工智能浪潮及其对6G的影响研究报告.pdf

V 1.0ngmn.orgAI SURGE AND ITS IMPLICATIONS FOR 6G2AI SURGE AND ITS IMPLICATIONS FOR 6Gby NGMN AllianceVersion:1.0Date:19 February 2026Document Type:PublicProgramme:6GApproved by/Date:NGMN Board,16 February 2026Public documents(P):2026 Next Generation Mobile Networks Alliance e.V.All rights reserved.No part of this document may be reproduced or transmitted in any form or by any means without prior written permission from NGMN Alliance e.V.The information contained in this document represents the current view held by NGMN Alliance e.V.on the issues discussed as of the date of publication.This document is provided“as is”with no warranties whatsoever including any warranty of merchantability,non-infringement,or fitness for any particular purpose.All liability(including liability for infringement of any property rights)relating to the use of information in this document is disclaimed.No license,express or implied,to any intellectual property rights are granted herein.This document is distributed for informational purposes only and is subject to change without notice.Readers should not design products based on this document.3CONTENTS EXECUTIVE SUMMARY 401 INTRODUCTION 602 IMPACTS OF AI TRAFFIC ON NETWORKS 7 2.1 Traffic Growth 7 2.2 Shift in Network Requirements803 NETWORK FOR AI 9 3.1 Performance vs.Business Value 9 3.2 Capabilities Beyond Connectivity 904 AI FOR NETWORK AND IMPLICATIONS FOR 6G ARCHITECTURE EVOLUTION 11 4.1 Network Management Layer 11 4.2 Core Network 12 4.3 Radio Access Network 12 4.4 Key Challenges and Considerations 12 4.5 Implications for 6G Network Architecture Evolution 1305 CONCLUSION&STANDARDISATION FOCUS AREAS15 5.1 Conclusion 15 5.2 Recommended Standardisation Focus Areas 1506 LIST OF ABBREVIATIONS 1707 REFERENCES 1808 FIGURES 19ACKNOWLEDGEMENTS 204EXECUTIVE SUMMARYThis is a pivotal moment in the telecommunications industry,propelled by the unprecedented AI surge and the beginning of 6G standardisation.AI is advancing at a rapid pace and will remain a dominant force,reshaping society far beyond the 6G era.This document consolidates NGMNs perspectives on how AI will likely impact 6G standardisation,providing guidance for ongoing 6G studies.This study examines three key dimensions:(1)impact of AI traffic on networks,(2)network for AI,and(3)AI for network and implications for 6G architecture evolution.Impact of AI Traffic on NetworksThe rapid proliferation of AI applications particularly those with autonomous,task-driven capabilities-introduces significant uncertainty into future network demand.While the precise impact of these AI-driven workloads on traffic patterns is difficult to predict,several factors could materially alter todays assumptions:Multi-modal AI applications:Services requiring real-time video exchange may drive substantial traffic growth and shift traditional traffic patterns.AI-enabled devices and use cases:Consumer applications(e.g.AR glasses)and enterprise scenarios(e.g.autonomous vehicles)could require frequent upload of images and video after local processing,increasing uplink demand and challenging current downlink-heavy network designs.Geographic and device density:AI-intensive areas and device clusters may experience sharp,localised surges,creating increasingly uneven traffic patterns.Given these uncertainties,network design must prioritise flexibility.Standards Development Organisations should explore mechanisms that allow semi-permanent adjustments in uplink/downlink ratio without requiring major standard revisions,as well as solutions to enhance the uplink coverage.This adaptability will be critical to accommodate evolving AI-driven requirements across diverse devices,networks and regions.Network for AI6G should go beyond providing connectivity services to deliver new AI enabled services and capabilities(e.g.new data exposure),by designing networks that are more intelligent,flexible,and trustworthy.Key design enablers include:Flexible(e.g.token-based)charging models reflecting real resource use.Dynamic and intelligent networking for AI agents collaboration.Support for explicit QoS and computing demand from an AI-based application,to facilitate meeting the required QoS at minimum cost and environmental impact.Enhanced QoS and adaptive policy control to support traffic routing achieving seamless connectivity.Unified data and model frameworks across devices and domains.Secure trust,authentication and authorisation mechanisms for AI agents digital identity.AI for Network and Implications for 6G Architecture Evolution AI is expected to be an important network capability for 6G networks,enabling more efficient usage of network resources,network automation,intent-based management and intelligent orchestration.AI could be applicable to all domains and different layers of the network,including the operation and maintenance.NGMN expects that 6G will be AI-ready,and the 5G Service-Based Architecture(SBA)will be considered as the starting point towards 6G architecture.5Challenges and considerations for adopting AI:Adoption of AI capabilities should allow agents and large language models(LLM)to be deployed in a way that avoids unnecessary impact on the existing architecture.This should not restrict the possible integration of AI-related features embedded within network functions(NFs).AI interfaces(e.g.,A2A,MCP)will complement existing and future APIs,ensuring readiness for the traffic volumes and capabilities required by emerging AI services.Multi-vendor interoperability frameworks are needed to ensure secure,scalable,and open ecosystems.Deployment strategies must align with cost and sustainability goals,and validation of real-world performance gains is essential.Continued support for non-AI alternatives if these alternatives are necessary to ensure reliability,flexibility and openness.Coordinated UEnetwork operation is needed,i.e.,to efficiently execute AI models in both two-sided and one-sided models.Recommended Standardisation Focus Areas Standardised architecture,protocols,and interfaces enabling efficient end-to-end support of AI functionalities,integrated across all domains(RAN,Core,Transport)and all network layers,including devices.Standards that support explicit network QoS and computing demand from an AI-based application,to facilitate meeting the required QoS at minimum cost and environmental impact.Standards that allow adaptability to support changing traffic patterns,accommodating uncertainty in the impact of evolving AI use cases.Evolution of the existing(5G SBA)network architecture should be justified by value driven AI use cases and service scenarios,ensuring alignment with societal and business needs.6G standards that support agent-to-agent and agent-to-network communications.Functional and performance requirements for AI capabilities across the 6G system.Establishment of interoperability and trust frameworks to enable secure,multi-vendor,and multi-agent deployments and operations(including models retraining,fine tuning).Emphasis on the reuse,adoption,or enhancement of“AI interface”from telco and non-telco worlds where appropriate and mainstream.(e.g.(A2A)Agent-to-Agent or(MCP)Model Context Protocol).601 INTRODUCTIONThe rapid evolution of large-scale AI models is driving a paradigm shift toward an“AI-native”era.The proliferation of large language and multi-modal models is enabling the emergence of AI agentsautonomous,collaborative,and self-learning entities that may outnumber human users in upcoming years.This shift toward pervasive,agent-driven ecosystems will fundamentally reshape industries,services,and everyday life.To support this transformation,networks may need to progressively introduce AI features for intent-driven programmability,autonomous operation,and dynamic compute distribution across central and edge domains.This evolution aims to deliver differentiated connectivity,high reliability,energy efficiency,and simplified operation,positioning 6G as the best network for AI and a foundation for AI-based applications,management,and innovation.As 6G standardisation enters a critical phase,the growth of AI and AI agents presents both opportunities and challenges for mobile network operators(MNOs).NGMN has outlined key 6G objectives and architectural design principles emphasising innovation across networks,AI,computing,sensing,modularity,operational simplicity,sustainability,trustworthiness,cloud nativeness,network-as-a-service,automation,smooth migration,and a disaggregated multi-vendor approach.These principles aim to guide the evolution of networks that are efficient,sustainable,cost-effective,and socially beneficial 1234567.To address the implications of AI on future network design and ensure alignment with NGMNs objectives,this document examines three dimensions from an operators perspective and highlights recommended standardisation focus areas to support industry alignment:Impact of AI traffic on networks Network for AI AI for network and implications for network architecture evolution702 IMPACTS OF AI TRAFFIC ON NETWORKS2.1 TRAFFIC GROWTHToday,mobile data consumption is dominated by video applications,accounting for 70-75%of total traffic 89.A handful of social media and streaming services contribute more than 50%of this demand.Although AI applications have grown exponentially,their current impact on mobile network traffic remains modest with primary interactions being text-based 10.This could change as AI services proliferate,but predicting the scale of impact remains highly speculative due to several factors:Optimisation of AI Models AI models are being optimised using techniques such as quantisation,pruning and reduced token sizes to enable efficient high-performance inference directly on device.11 Local Processing More complex AI models are expected to run natively on device as chipsets evolve with larger and more capable Neural Processing Units(NPU),faster on-chip memory and cache,increased RAM allocation for AI workloads and tighter integration of hardware with AI frameworks and runtime engines.Unclear Adoption Curve End-user adoption curve:it remains uncertain which new services provide true additional value for end-users,impacting services adoption,traffic curves and commercial models.Regulatory and Privacy Constraints Several challenges would need to be resolved,for data-heavy AI features,such as automatic image or video capture via AR glasses.Against this uncertainty,the potential impact of AI on traffic growth needs to be considered in the following aspects:Substitution of Current Demand Multi-modal AI applications are likely to proliferate capturing more user attention,with smartphones likely remaining a primary interface.However,it is expected that most video traffic from these applications will replace existing user behaviour such as watching social media video feeds rather than creating truly incremental demand.Potential Rise of Wearables AR glasses and similar interfaces could dramatically increase traffic if they continuously interact with cloud-based AI applications using video or images.This traffic would be considered incremental,rather than substitutional,but adoption hinges on overcoming privacy and security concerns as discussed above.Enterprise and Other Applications Autonomous drones,connected cars,humanoid robots/cobots and industrial AI use cases could add significant trafficprovided technological and regulatory hurdles are cleared.Uplink Trends Current uplink demand is moderate,but future use cases such as AI agents could reverse this trend 10.AI agents with advanced perception and reasoning capabilities may reside on smartphones or wearables,continuously gathering data and interacting autonomously-potentially generating far more data than humans,subject to battery capacity and computational power of the device.However,this shift is uncertain,as many AI agents could instead operate in the cloud,performing inference and delivering recommendations to the user.Future scenarios differ greatly in both likelihood and scale of impact.Use cases that drive truly incremental video traffic beyond todays demand will exert the greatest pressure on networks.While some scenarios 8present significant potential for increased demand,they must be weighed against their likelihood when setting priorities for network evolution.This uncertainty makes flexibility a cornerstone of 6G standardisation ensuring the network can adapt seamlessly to diverse and unpredictable requirements.2.2 SHIFT IN NETWORK REQUIREMENTSThe rise of AI may introduce fundamental changes in both the form and direction of traffic:Machine-oriented Media Traditional networks primarily carry human-perceivable content(text,images,audio,video).In contrast,agent-to-agent communication may involve exchanging models,feature vectors,latent representations,and other forms of information optimised for machines rather than humans.Uplink-heavy Behaviour While todays traffic is mostly downlink-dominated,many AI-enabled use cases are assumed to reverse this pattern.For instance,AR glasses with AI may require continuous uplink transmission of environmental images,and AI-inferenced autonomous vehicles may upload real-time video and sensor data more often,in contrast to traditional connectivity patterns.6G networks should support these use cases with sufficient flexibility to increase uplink traffic as a major design driver for 6G networks.For example,increased uplink(UL)slot occurrences that maximise the UL transmission opportunities to manage the increased UL traffic expected with new services.Some of the proposals that are being discussed in industry and under review in 3GPP are around the definition of flexible and dynamic downlink(DL)/UL patterns,for example,Full-Duplex or Sub-band Full Duplex operation.Enhancing UL coverage is also a desirable feature.Regional and Sectoral Variability The impact of AI traffic will differ across regions and industries.Urban centers are likely to experience more AI traffic surges than rural or remote areas.Certain industries such as manufacturing,transportation,healthcare,and smart cities may generate higher volumes of AI traffic.AI-intensive areas and device clusters may experience sharp,localised surges,creating uneven traffic patterns.903 NETWORK FOR AIAI-driven applications impose new requirements on 6G networks,encompassing not only improved connectivity performance but also new capabilities beyond connectivity.3.1 PERFORMANCE VS.BUSINESS VALUEFor performance improvements related to traditional connectivity,it is essential to validate the necessity of any network enhancements from a business value perspective in order to avoid unnecessary investment and resource waste.While network optimisation can improve user experience to some extent,not all scenarios require extreme performance gains,as the existing services offered may not be directly impacted by these network enhancements.For example,humans generally have a relatively high tolerance for latency